What if you were starting a firm today?
Part 9 of The Self-Improving Firm — a nine-part series on what AI-native looks like for a UK accounting firm
Daniel Lawrence, CEO bots for that
Imagine a Tuesday morning in September, next year.
A partner at a Top 50 firm has spent the weekend not sleeping. Twenty-two years at the firm. Made partner at thirty-four. Runs a portfolio of clients they mostly like, in a firm they mostly respect, in a role they have mostly grown into. The pay is good. The pension is better. The route from here to retirement is clear enough that they could sketch it on the back of an envelope.
On Monday afternoon, they hand in their notice.
On Tuesday morning, they sit at their kitchen table with a laptop, a notebook, and a cup of coffee going cold, and they begin the work of designing a firm from scratch. No inheritance. No committee. No compromises made to keep other partners or investors happy. No practice management system chosen because it was the least bad option in 2014. No office lease. No twenty-year-old methodology bolted onto Xero. Nothing.
Just one question. What would they actually build?
This is the thought experiment the series ends with. It is not a hypothetical. Partners are doing this right now, quietly, at kitchen tables across the country. The firms they build over the next five years will be the ones the current mid-tier firms are competing with for the next thirty. It is worth being specific about what those firms are going to look like.
The staffing model, on day one
The partner starts with themselves.
Not with a plan for how many junior accountants they will hire in year one. Not with an org chart. With one person and a laptop.
They know, as I’ve pointed out in this series, that the compliance production line is largely automatable. They know that a serious AI-assisted working papers process can produce first-pass output at a fraction of the human cost. They know that the coordination layer they have grown up with is dissolving. They know that the partner economics of an AI-native firm are inverted: fewer clients per partner, deeper time on each, partners as high-density human interfaces.
So the first hire is not a junior. It is the infrastructure.
They build, or buy, or configure, the intelligence layer first. The sensor layer that captures every client signal. The decision layer that codifies what the firm will and will not do. The closed-loop system that ensures the firm learns from its own work. This takes them, say three months. During that time, they are not billing. They are building the thing that will make the billing possible.
Their first human hire is not a trainee. It is a Player-Coach. A senior, technically-capable accountant who wants to work in the new model and who is willing to be the first person alongside the partner in the flow of engagements. Together, the two of them can hold more clients than the partner alone could, because the infrastructure is doing the work that used to require a team of five.
Their second hire, probably around the fourth to sixth month, is a DRI. A partner-track individual who will own a subset of the firm’s client outcomes end to end. The firm is now three people plus the infrastructure. Compare this to the traditional firm at a similar revenue level, which would have needed perhaps twelve to fifteen people, most of them juniors doing compliance and preparation work.
The junior hires come later, if they come at all. And when they come, they are not preparing files. They are reviewing AI-prepared files. They are sitting in on the partner’s client conversations. They are learning to recognise good work by seeing a great deal of it, and they are learning to be accountants by being present when partners are being accountants. The training pathway is not the legion pathway. It is an apprenticeship model that ceased decades ago and is now, quietly, reinventing.
Year two, the firm has perhaps eight to ten people. It is running a client base that a traditional firm would need thirty people to service, and it is running it more profitably, at a lower price point to the client, with better staff retention because the work is more interesting.
This is not an unrealistic projection. It is the operational logic of the previous eight parts of the series, followed through.
The service model, on day one
The partner does not build a compliance factory that also does some advisory work when clients ask.
They build the inverse. An advisory firm that produces compliance as a by-product.
The distinction matters. In a traditional mid-tier firm, compliance is the anchor tenant. It is what the firm sells, what the fees are calculated against, what the staff are structured around. Advisory is bolted on the side, sold to willing clients, often underpriced because nobody is quite sure what to charge.
In the AI-native firm, compliance is the substrate. It is produced by the infrastructure, at low marginal cost, as a natural consequence of holding the client’s data and knowing the client’s business. The firm’s revenue is not primarily compliance revenue. It is advisory revenue, relationship revenue, judgement revenue, produced by partners spending their high-density human time on the work that only partners can do.
This changes the fee model. Fixed monthly fees, priced for the relationship rather than the file, priced against outcomes rather than hours. Advisory work priced against the value it creates, not against a percentage uplift on the compliance base. Some client segments, particularly the ones who want compliance-only, at a price, and no relationship, may find themselves better served by other firms. That is fine. The AI-native firm is not trying to serve everyone. It is trying to serve the clients for whom the partner’s time is worth what the firm is charging for it.
This has an interesting implication. The AI-native firm is likely to be smaller in client count than the equivalent-revenue traditional firm, and larger in revenue per client. The economics are fundamentally different. So is the client conversation. So is the marketing. So is the recruitment pitch to the partner-track hires the firm needs. Everything reconfigures around the observation that compliance is no longer the product.
The client model, on day one
The partner has to make an early decision about which clients they are building the firm for.
The traditional mid-tier firm serves a mix. Sole traders and micro-businesses at the bottom of the client base, SMEs in the middle, owner-managed businesses and mid-market clients at the top. The mix is inherited. It is what the firm has grown into. Each client segment has different economics, different service expectations, different sensitivities to fees, different capacity to benefit from advisory work.
The AI-native firm cannot serve all of them at once, at least not on day one. The infrastructure investment does not pay back if it is deployed against the wrong client segment.
The right segment for most AI-native firms is the middle. Owner-managed businesses (“OMB”). Companies between five and fifteen million in revenue. Clients who are big enough to benefit from advisory work but not so big that they employ their own experienced finance experts. Clients who value relationship, who make decisions that turn on judgement, and who will pay for both. Clients for whom the compliance is a hygiene factor and the strategic conversation is the thing they actually want.
This is not the only viable segment. High-net-worth private clients are another. Specialist sectors are another. But the mid-market OMB is the segment where the AI-native firm’s economics work most cleanly, because the client is sophisticated enough to value the model and small enough to be genuinely served by a small team. The client’s who are building their own AI-native businesses.
The partner picks a segment. They design the firm for it. They resist the pull, familiar to every partner who has ever grown a firm, to say yes to work that does not fit, on the grounds that the fee is welcome. The AI-native firm scales through depth in a segment, not through breadth across segments.
Drowning in manual, repetitive work? Tell us the task and we’ll show you what to automate.
The cultural model, on day one
Here is the part of the thought experiment that partners typically do not think about, and that is arguably the most important.
The partner has to decide, on day one, what the firm is going to feel like to work in.
The traditional mid-tier firm has, over decades, accumulated a culture. Some of it is good. Some of it is habit. Some of it is inherited from partners who left twenty years ago and whose views were formed in an entirely different era. The culture is rarely designed. It emerges.
The AI-native firm has the opportunity, one time, at the beginning, to design the culture deliberately. Some choices to make explicitly.
Errors are recorded honestly and used for improvement rather than punishment. The closed-loop system requires this, and the choice has to be made in the first month, because it is very difficult to introduce after fear has been established.
Judgement is aggregated deliberately. Partner conversations are actually about difficult calls, not about status updates. Written reflections from senior people are captured and read. Mentorship is treated as senior work, not junior overflow.
The DRI model is the operating norm. Every client outcome has one accountable person. Shared responsibility is treated as an anti-pattern, not a comfort.
Partners are Player-Coaches by design. They are not permitted to become pure managers. If they stop opening files, stop taking client calls, stop being present in the flow of engagements, they lose something the firm cannot afford to lose.
Technology is treated as ephemeral. The specific tools will change. The specific vendors will change. The specific integrations will change. What matters is the domain knowledge the firm accumulates, in a form the firm owns, that survives any specific technology stack. This is the point the source material made most forcefully and that the profession has heard least clearly.
Client relationships are treated as the firm’s asset, but their portability is respected. Partners who leave take their clients with them, because the trust is genuinely personal. The firm designs its economics around this rather than pretending it is not true.
None of this is optional in the AI-native firm. Each of these cultural choices is what makes the infrastructure worth building. A firm that installs the sensor layer but keeps the political-blame culture will produce technically closed-loop, operationally open-loop, output that gets worse over time rather than better.
Why this matters for firms that already exist
The point of the thought experiment is not that every partner should leave their firm and start over. Most will not. Most should not.
The point is that the partner at the kitchen table, in September next year, is the implicit benchmark. Every current mid-tier firm is now competing, whether it knows it or not, with the version of itself that a serious partner would build from scratch. Not against the Big Four. Not against the bookkeeping consolidators. Against the firm a good partner would build, deliberately, with what is now available.
That benchmark is not distant. The partner at the kitchen table exists. There are, right now, in this country, dozens of them. Some of them are already in their second year. Some are still in the notice period. The firms they build will not be enormous. They will not need to be. They will be small, dense, technically ambitious, culturally deliberate, and they will serve the exact clients the mid-tier firms would most like to keep.
The firms that stay competitive against this benchmark will be the ones that treat the question seriously. Not as a strategic-away-day exercise. As the operational logic of every decision the firm makes for the next five years.
The firms that do not will find, at the end of those five years, that the partners they hoped would drive their transformation have left to build something better. And that the clients they hoped would tolerate a slower pace of change have been quietly poached by a firm they had not previously heard of, operating with a fraction of their overhead, holding better data on their clients than they do, and making the partners’ phones ring more often than the mid-tier firm ever did.
A closing observation
The Roman legion, over centuries, defeated most of what it faced. It also, eventually, lost.
It did not lose because its soldiers were less brave, its centurions less experienced, or its manuals less refined. It lost because its opponents worked out that the legion’s shape was designed for a world that had already changed. Heavy cavalry that outmanoeuvred the infantry. Decentralised raiding that outlasted the supply lines. Political fragmentation that starved the recruitment pipeline. Each of these was a challenge the legion, in principle, could have adapted to. Some of them, in practice, it did adapt to. Others, it did not. The ones it did not adapt to are the ones that ended it.
The accounting firm has not yet had to face its opponent. The technology has arrived. The economics are shifting. The strategic urgency, as Part 6 argued, is compounding. But the firm the current mid-tier will find itself competing against is not yet visible from where the current mid-tier is sitting. It will be visible in two to three years. It will be everywhere in five.
The partner at the kitchen table is not a threat to the mid-tier firm. They are a mirror. They are what the mid-tier firm could be, if it had the courage to design itself deliberately, with the constraints removed and the assumptions questioned. The mid-tier firm has an advantage the kitchen-table partner does not: scale, capital, client trust, established relationships. If the mid-tier firm uses those advantages to build something like what the kitchen-table partner is building, faster than the kitchen-table partner can grow to challenge it, the mid-tier firm wins.
If the mid-tier firm uses those advantages to defend the shape it already has, it will find that the advantages, one by one, stop being advantages. The scale becomes overhead. The capital becomes committed cost. The client trust erodes as clients see better-run firms in action. The established relationships walk out with the partners who built them.
This is not a prediction. It is a description of a decision the mid-tier firm’s leadership is making, right now, whether they know it or not.
The question the series ends on is the same question it started with, in a different form. In Part 1, the question was whether the pyramid was still the right shape for a firm to be. In Part 9, the question is what shape a serious partner would build if they were starting today, and whether the firm you currently run is closer to that shape or further from it than it was a year ago.
If it is closer, keep going. The work is difficult and it is not glamorous and the returns compound over time.
If it is not closer, the work of the next five years is not the same as the work of the last five. It is deeper, harder, and more honest. It is a redesign, not a refresh. It is a conversation with your partners you have probably been avoiding. It is a set of decisions about your firm that you have inherited rather than made.
The Roman legion was extraordinary for six centuries. It was replaced by what came after. The question is not whether change is coming to this profession. The question is whether the firm you have built will be part of what comes next, or part of what came before.
That is the question the series has been asking all along. It is the question every partner in the profession should be asking, out loud, this year.
The answer, whatever it is, is worth having.
Daniel Lawrence is the CEO and co-founder of bots for that and creator of the Automation Operating System (AOS). He has spent more than a decade deploying enterprise automation and AI in regulated industries including accounting and professional services. The Self-Improving Firm is a nine-part series exploring what AI-native operations look like for mid-tier and large UK accounting firms.
